Archives of the TeradataForum

Message Posted: Wed, 03 Mar 2010 @ 14:58:51 GMT

	<Prev	Next>		<<First	<Prev	Next>	Last>>

Hi Clay,

Barry has already given some ideas and thoughts. To add to those...

Have you considered splitting your one fact table into two physical tables each using the same PI ? If most queries go after a subset of the columns then that will reduce the physical size of the table thus reducing the impact of skewed data.

Another way may be to use PPI. If your WHERE clause (in most or all 300 cases ?) includes for instance a date column, partition on that column so that although the data is still skewed across the AMPs, within an amp the number of rows 'lumped together' in a single row hash is reduced because of the partitioning.

Barry mentioned using a JI. Using a single-table JI (or even a sparse JI ?) might not give you single-amp access but it might give you enough to achieve your business processing goals.

Cheers,

Dave

Ward Analytics Ltd - Information in motion ( www.ward-analytics.com )

	<Prev	Next>		<<First	<Prev	Next>	Last>>